The longitudinal relations between teacher-student relationship quality (TSRQ) and student achievement were examined to determine the directional associations between the quality of teacher-rated closeness and conflict with students, and measured math and reading achievement in a large, multisite sample of U.S. youth at first, third, and fifth grade. A longitudinal confirmatory factor analysis model of panel data was employed. After testing longitudinal factorial invariance across time, we tested heterogeneity in the factor variances and differences in the latent means. Math and reading achievement had longitudinal reciprocal relations. Math achievement explained small differences in subsequent teacher-rated closeness after controlling for previous levels of math achievement and teacher-rated closeness. Teacher-rated conflict served as a small but significant predictor of subsequent math achievement across measured time points but previous teacher-rated closeness did not explain subsequent reading or math achievement at any time point. Teacher-rated conflict was relatively stable across grades, whereas teacher-rated closeness varied from first to third grade. Reading and math achievement were highly stable predictors of future achievement. The findings suggested that in a lower-risk sample, measures of TSRQ and achievement may serve as predictors or outcomes and directionality of effects should not be assumed in advance.